System and method for facilitating prediction data for device based on synthetic data with uncertainties

Abstract

Embodiments described herein provide a system for facilitating a training system for a device. During operation, the system determines a system model for the device that can be based on empirical data of the device. The empirical data is obtained based on experiments performed on the device. The system then generates, from the system model, synthetic data that represents behavior of the device under a failure. The system determines uncertainty associated with the synthetic data and, from the uncertainty, determines a set of prediction parameters using an uncertainty quantification model. The system generates training data from the synthetic data based on the set of prediction parameters and learns a set of learned parameters associated with the device by using a machine-learning-based classifier on the training data.

What is claimed is: 
     
       1. A computer-implemented method for facilitating a training system for a device, the method comprising:
 determining, by a computer, a system model for the device based on empirical data of the device, wherein the empirical data is obtained based on experiments performed on the device, and wherein the system model outputs data that indicates how the device operates for a set of actions performed on the device; 
 generating synthetic data by applying, on the system model, one or more actions representing behavior of the device under a failure; 
 determining a range of parameters indicating uncertainty associated with the synthetic data; 
 determining a set of prediction parameters from the range of parameters using an uncertainty quantification model to reduce parameters associated with the uncertainty; 
 generating training data from the synthetic data based on the set of prediction parameters; and 
 learning a set of learned parameters for a prognosis of the device using a machine-learning-based classifier or a regression model on the training data. 
 
     
     
       2. The method of  claim 1 , wherein the uncertainty quantification model is based on generalized Polynomial Chaos (gPC) expansion. 
     
     
       3. The method of  claim 1 , wherein the machine-learning-based classifier is based on one or more of:
 an optimization of a set of input parameters to the system model; and 
 a state estimation of the set of input parameters to the system model. 
 
     
     
       4. The method of  claim 3 , wherein the state estimation is based on a regression model using one or more of: a Kalman filter, an extended Kalman filter, and a particle filter. 
     
     
       5. The method of  claim 1 , wherein the reduction of parameters associated with the uncertainty mitigates uncertainty, which is indicated in the range of parameters, in the synthetic data by reducing computation overhead. 
     
     
       6. The method of  claim 1 , further comprising:
 determining a mode of operation of the device that represents the failure of the device; and 
 determining the prediction parameters for the mode of operation. 
 
     
     
       7. The method of  claim 1 , further comprising:
 determining prognosis information for the prognosis of the device from the set of learned parameters; and 
 determining a prognosis policy for the device based on the prognosis information. 
 
     
     
       8. The method of  claim 1 , further comprising:
 determining a current state of the device based on a current environment for the device, wherein the current environment corresponds to the failure of the device; and 
 determining an operation corresponding to the current state based on the set of learned parameters. 
 
     
     
       9. A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for facilitating a training system for a device, the method comprising:
 determining a system model for the device based on empirical data of the device, wherein the empirical data is obtained based on experiments performed on the device, and wherein the system model outputs data that indicates how the device operates for a set of actions performed on the device; 
 generating, from the system model, synthetic data by applying, on the system model, one or more actions representing behavior of the device under a failure; 
 determining a range of parameters indicating uncertainty associated with the synthetic data; 
 determining a set of prediction parameters from the range of parameters using an uncertainty quantification model to reduce parameters associated with the uncertainty; 
 generating training data from the synthetic data based on the set of prediction parameters; and 
 learning a set of learned parameters for a prognosis of the device using a machine-learning-based classifier or a regression model on the training data. 
 
     
     
       10. The computer-readable storage medium of  claim 9 , wherein the uncertainty quantification model is based on generalized Polynomial Chaos (gPC) expansion. 
     
     
       11. The computer-readable storage medium of  claim 9 , wherein the machine-learning-based classifier is based on one or more of:
 an optimization of a set of input parameters to the system model; and 
 a state estimation of the set of input parameters to the system model. 
 
     
     
       12. The computer-readable storage medium of  claim 11 , wherein the state estimation is based on a regression model using one or more of:
 a Kalman filter, an extended Kalman filter, and a particle filter. 
 
     
     
       13. The computer-readable storage medium of  claim 9 , wherein the reduction of parameters associated with the uncertainty mitigates the uncertainty, which is indicated in the range of parameters, in the synthetic data by reducing computation overhead. 
     
     
       14. The computer-readable storage medium of  claim 9 , wherein the method further comprises:
 determining a mode of operation of the device that represents the failure of the device; and 
 determining the prediction parameters for the mode of operation. 
 
     
     
       15. The computer-readable storage medium of  claim 9 , wherein the method further comprises:
 determining prognosis information for the prognosis of the device from the set of learned parameters; and 
 determining a prognosis policy for the device based on the prognosis information. 
 
     
     
       16. The computer-readable storage medium of  claim 9 , wherein the method further comprises:
 determining a current state of the device based on a current environment for the device, wherein the current environment corresponds to the failure of the device; and 
 determining an operation corresponding to the current state based on the set of learned parameters. 
 
     
     
       17. A computer system; comprising:
 a storage device; 
 a processor; and 
 a non-transitory computer-readable storage medium storing instructions, which when executed by the processor cause the processor to perform a method for facilitating a training system for a device, the method comprising:
 determining a system model for the device based on empirical data of the device, wherein the empirical data is obtained based on experiments performed on the device, and wherein the system model outputs data that indicates how the device operates for a set of actions performed on the device; 
 generating, from the system model, synthetic data by applying, on the system model, one or more actions representing behavior of the device under a failure; 
 determining a range of parameters indicating uncertainty associated with the synthetic data; 
 determining a set of prediction parameters from the range of parameters using an uncertainty quantification model to reduce parameters associated with the uncertainty; 
 generating training data from the synthetic data based on the set of prediction parameters; and 
 learning a set of learned parameters for a prognosis of the device using a machine-learning-based classifier or a regression model on the training data. 
 
 
     
     
       18. The computer system of  claim 17 , wherein the uncertainty quantification model is based on generalized Polynomial Chaos (gPC) expansion; and
 wherein the machine-learning-based classifier is based on one or more of: 
 an optimization of a set of input parameters to the system model; and 
 a state estimation of the set of input parameters to the system model. 
 
     
     
       19. The computer system of  claim 17 , wherein the reduction of parameters associated with the uncertainty mitigates the uncertainty, which is indicated in the range of parameters, in the synthetic data by reducing computation overhead. 
     
     
       20. The computer system of  claim 17 , wherein the method further comprises:
 determining a mode of operation that represents the failure of the device; and 
 determining the prediction parameters associated with the mode of operation.